Hi @binlogreader , how are you doing today?

Your understanding looks right to me. I would use streaming tables when the use case involves CDC, SCD, multiple flows, or continuous record processing. I would use materialized views when the main goal is to always maintain the latest result from upstream data. Since both are pipeline-owned, I would not make ownership the main deciding factor. For keyed upserts, AUTO CDC into a streaming table still feels simpler and more reliable. For sharing data across teams, I would prefer exposing a stable Delta table or governed data contract instead of making another team depend directly on the pipeline design.

Hope this helps!